A nonhomogeneous hidden Markov model (NHMM) is used to stochastically simulate summer (June–August) daily precipitations in the middle and low reaches of the Yangtze River in Eastern China, with driving forcing from three global climate models (GCMs). Simulations cover the historical period from 1961 to 2005 and from 2006 to 2100 following the RCP4.5 scenario. The model is first evaluated against data from the regional observation network. The results show that NHMM effectively enhances the ability of GCMs in simulating summer daily rainfall in the region. For future projection at different time horizons of the 21st century, the spectral distribution of regional precipitations (in function of their intensity) shows consistent changes with a decrease of occurrence probability for light rain (<10 mm/day) and an increase for heavy rain (>10 mm/day). Among variables of interest, total precipitation (PRCPTOT), precipitation intensity, the number of rainy days for daily precipitation exceeding 10 mm (R10mm) and 95th percentile of precipitation (P95), all show a gradually increasing trend in the 21st century, and geographically an eastward gradient with smaller increase (or even weak decrease) for the west and larger increase for the east. It is noted that obvious changes occur in the eastern region with 95% significance level, and PRCPTOT or R10mm increases by 40–60% in the late 21st century. Further quantitative assessment is performed for global warming of 1.5 and 2°C. The half‐degree additional warming makes R10mm change by −3.7, 2.4 and 12.1% over western, central and eastern regions, respectively.